Starlings in flight Starlings in flight understanding patterns of - - PowerPoint PPT Presentation

Irene Giardina

INFM-CNR, Department of Physics, Rome La Sapienza ISC Institute for Complex Systems, CNR Rome

and

Starlings in flight Starlings in flight

understanding patterns of animal group movements understanding patterns of animal group movements from the complex system perspective from the complex system perspective

European STREP project STARFLAG STARFLAG

Pavia, November 2005

Termini railway station, Rome Evening roosting time, November 2004 STARLING FLOCKS

Collective phenomena Collective phenomena often occur in biological systems

Bacteria colonies , blood cells, insects swarms, fish schools, birds flocks, quadrupets herds

What are the rules governing coordination and collective motion ?

Collective phenomena have been widely studied in physics

• Cooperative behaviour in phase transitions and ordering
• Local interactions can generate long range order
• Universality, renormalization the details are not important
• Efficient simple models

the microscopic mechanisms determining flocking pattern formation and coordinated collective motion are local and simple and do not depend dramatically on the complex nature of the individuals

The Physics Paradigm

Hypothesis !

SIMPLE MODELS

Minimal Models of Flocking

Each individual bird determines its direction of motion on each time step by averaging Each individual bird determines its direction of motion on each time step by averaging the direction of its neighbours (allelomimesis) with some noise the direction of its neighbours (allelomimesis) with some noise Self-Propelled Particles

Reynolds, Comput. Graph., 87 Vicsek et al., PRL 95

direction velocity noise

• Nonequilibrium analog of the ferromagnetic XY model (in 2D) Rotational Symmetry
• Onset of collective motion for small noise, even in 2D (Mermin-Wagner does NOT hold)
• Navier-Stokes like equations for the coarse-grained velocity Toner & Tu, PRL 1995

Convective relevant non-linear terms Non trivial RG fixed point Exact exponents in D=2 Effective long-range interactions

SPP with cohesion

Gregoire, Chate & Tu, PRL 2001 Gregoire & Chate, PRL 2004 Gregoire, Chate & Tu, PRE 2004

Hard-core repulsion + Short-range attraction ( r0 )

• Non trivial infinite space limit

cohesive moving flocks in infinite space

• Complex phase diagram

solid moving solid liquid moving liquid

• Discontinuous first order transitions

Moving/Non moving Cohesive/Sparse Gas - Liquid - Solid

Experiments

Stereoscopic 3D reconstruction of

Flock shape and movement Individual birds positions Individual birds trajectories

Stereoscopic Photography

• image elaboration
• birds recognition
• stereoscopic matching
• epipolar post-calibration

Stereometry

2 D images 3D coordinates

xB xA

Z f d

lens CCD CCD lens

A B

Stereoscopic shift Stereoscopic shift

• the larger the distance, the better the resolution

f = 4000 Z = 200 m δZ = 0.5 m δs = 1

d = 20 m

neighbouring birds

• misalignements strongly affect absolute distances

α = 0.001 rad δZ/Z = 2.0 / 200 α = 0.0003 rad δZ/Z = 0.5 / 200

α

The setup

Palazzo Massimo, Rome 5fps @ 16m/s : birds travel 3.2 m between two consecutive shots 25 m Syncronized ( < 1 ms) interlaced

• Alignement

There are 5 external angles to be fixed 4 with precision ~ 0.001 rad 1 with precision ~ 0.0005 rad

• Calibration

There are the internal parameters to be calibrated (or postcalibrated)

• Temporization

2 interlaced cameras 10 fps 1.6 m

Matching and 3D reconstruction

• Bird recognition

Contrast filters, segmentation algorithms

• Matching

1) Pattern recognition algorithm Zero-matching (partial) 2) Epipolar geometry F = fundamental matrix 3) K-assignement Matching between different set of points with measure F M0

depends on the angles ! affected by JPEG Noise

Planar structure ! Planar structure !

A more complex flock

Discoidal shape Discoidal shape

r

flock bio

* +

g(r) r Γ(r)

Radial distribution function g(r) Liquid like !!!

Conditional mass Γ(r)

Scale free (???) very preliminary Density = N/V

Synthetic flock with same V, N and overall shape as the bio one, but with a uniform distribution of points

Finite size effect (L< L*) Errors in segmentation Errors in matching ?

Summary and Perspectives

• 3D reconstruction of starling flocks is demanding but

possible

• Experimental efficiency related to

Camera specifications (Canon Eos D Mark II, 8.2 Mp, 8.5 fps) alignement capabilities

• Static reconstruction of individual flocks
• Statistics correlation functions, shape, heterogeneity
• Dynamics trajectory reconstruction, diffusion, convection

Comparison with models

Alberto Orlandi

(INFM-CNR, STARFLAG postdoc)

computer vision, epipolar geometry

Vladimir Zdravkovic

(INFM-CNR, STARFLAG postdoc)

experimental setup, data taking

Michele Ballerini

(INFM-CNR, STARFLAG graduate student)

electronics, timer, data taking

Evaristo Cisbani (ISS)

electronics, timer

Nicola Cabibbo (La Sapienza) Andrea Cavagna (INFM & ISC-CNR ) Irene Giardina (INFM & ISC-CNR) Giorgio Parisi (La Sapienza, INFM & ISC) Andrea Procaccini

(La Sapienza, PhD student) experimental setup, data analysis

Massimiliano Viale (Roma 3 & INFM-CNR, PhD student)

epipolar geometry

The team